Exploring the Effect of Menstrual Hygiene Management Spaces on Academic Achievement: Evidence From Young Lives' School Survey in Ethiopia

(1)

EXPLORING THE EFFECT OF MENSTRUAL HYGIENE MANAGEMENT SPACES ON ACADEMIC ACHIEVEMENT:

EVIDENCE FROM YOUNG LIVES’ SCHOOL SURVEY IN ETHIOPIA

Jessica Huynh

Thesis submitted in partial fulfilment of the requirements for the degree

Master of Philosophy in Global Development Theory and Practice, with specialization in Health Promotion

Spring 2019

Department of Health Promotion and Development Faculty of Psychology

University of Bergen

(2)

I

Acknowledgements

To my supervisors, Helga Urke and Paul Kellner. Thank you for your patience, support, and encouragement. You have challenged me to think further and I have learned and grown as a researcher because of you both. My gratitude also extends to Maurice Mittlemark. Thank you for your guidance and feedback on my analyses. I am truly grateful to have connections to knowledgeable mentors who have helped with the strategic development of this thesis.

To my family. My husband, Lars Andreas Selberg, you will always be my beacon of light, reminding me that every cloud has a silver lining. My parents-in-law, Anita and Aksel Selberg, your kindness and generosity means the world to me. My sisters, Stacey, Jackie, Julia, and Shannon Huynh, you have been my source of comfort and encouragement when the road was tough. From the bottom of my heart, thank you.

To my friends: Steffy Earnest, Brandon Shih, and Jonathan Culich. You are my equals, you are my betters, and I am so appreciative of your continued support of my goals in life.

To my professors and colleagues in the master’s program. Thank you for sharing your

knowledge and experiences with me. You have enriched and broadened my perspective of the world and have significantly contributed and inspired me in my work.

(3)

II

Abstract

Background: Ethiopian schools require improved school environments in order to improve educational outcomes. Evidence suggests that a lack of sanitation spaces, specifically private areas to for girls to change and manage menstruation hygienically, comfortably, and with dignity, leads to school absenteeism, distraction, and disengagement.

Objective: The objective of the study was to explore the effects of menstrual hygiene management (MHM) spaces (a private place to wash menstrual rag and/or a place where female student can wash themselves privately) on math and English achievement scores and the extent in which environmental factors at the individual- , home-, and school/community levels can help explain those differences.

Data and Methods: 3844 adolescent girls, between the ages of 14 and 19 years, from the Young Lives 2016-2017 Ethiopian School Survey were included in the study. Math and English test scores were measured at Wave 1 to Wave 2. To account for the change in scores across time and the effect of MHM spaces on math and English scores, a mixed between- within subjects analysis of variance (ANOVA) was used. A mixed between-within subject analysis of covariance (ANCOVA) was used to assess the effects of MHM spaces on math and English scores while accounting for individual- , home-, and school/community factors.

Results and Discussion: The study found evidence that the availability of MHM spaces had a significant, yet very small effect on performance of math and English achievement tests in unadjusted analysis. However, adjusting for individual-, home-, and school- level covariates removed the effect between MHM spaces and achievement scores that were found in

unadjusted analysis. While the potential effects of MHM spaces on achievement tests in this study are small, other individual, family, and school characteristics measured in this study were found to be more important.

Conclusion: This study examined MHM spaces and its impact on academic achievement.

Though an important first step, providing MHM spaces does not, on its own, enable education for girls to fulfill its transformative potential. The potential for improving the Ethiopian girls’

education requires comprehensive consideration and interventions that operate on various environmental levels found in the ecological framework. Further research may expand on the findings of the study by improving the methodologies which include using direct

menstruation data, consideration of the quality of MHM spaces, and the use of multilevel linear modeling analyses.

(6)

V

Acronyms and Abbreviations

ANOVA – Analysis of variance ANCOVA – Analysis of covariance

EDRI – Ethiopian Development Research Institute ENLA – Ethiopian National Learning Assessment LMIC – Low and middle income country

MHM – Menstrual hygiene management

SPSS – Statistical Package for the Social Sciences TVET – Technical and vocational education and training

UNICEF – United Nations International Children’s Emergency Fund WASH – Water, sanitation, and hygiene

WHO – World Health Organization

(7)

1

1. Introduction

1.1 Background

Education plays a major role in the development of children, communities, and countries and is intrinsically linked to other development goals including the eradication of poverty,

empowerment, reducing hunger, and improved health (Herz & Sperling, 2004; Nunes, Lee, &

O’Riordan, 2016). However, across the developing world, millions of girls are not receiving the basic education needed to achieve these development goals. The United Nations

Educational, Scientific and Cultural Organization estimates that 264 million children between the ages of six to fourteen are not in school, nine million of which are girls living in sub- Saharan Africa (UNESCO, 2017). This is particularly worrying as education is critical to economic progress and global poverty reduction with broader implications for foreign policy improvements.

Poverty remains widespread in Ethiopia as a large portion of the Ethiopian population live in rural areas. With an economy that relies on agricultural production there are implications for the education sector. The country’s demographic makeup poses a problem of ensuring equitable access to education and the issue of relevance of the school curriculum (Federal Ministry of Education, 2015). Ethiopia’s demographic pressures have increased the demand of education, but it falls short. This can be seen in recent years as the country has made great advances in terms of increasing primary school enrollment, it still faces challenges in

retention, grade progression, and learning levels (Bow-Bertrand, Briones, & Favara, 2018).

From 2010 to 2015 net enrollment has increased from 73 percent to 85 percent, but attainment to higher grades are not apparent as many students leave the system early, reflected in a Grade 8 completion rate of only 47 percent (Ministry of Education, 2015; World Bank, 2019). The Ministry of Education (2015) cites that the share of female students at undergraduate level reached 32 percent in 2014/2015, but those that sat in the Grade 12 examination performed poorly. These stark figures suggest that Ethiopian policy makers should focus not just on increase enrollment but on improving conditions that enhance educational outcomes, such as providing essential facilities to meet the demands of increase enrollment.

There is evidence to suggest schools in Ethiopia lack the necessary facilities to meet the demands of increase enrollment. Inadequate school sanitation facilities have been cited to be a barrier to girls’ access to education. The growing anecdotal evidence reveals the gender

(8)

2

discriminatory nature of school environments as female students report challenges managing menstruation in the safety of private places, preventing their abilities to succeed (Herz,

Barbara; Sperling, Gene, 2004; Mason et al., 2013). Thus, research suggests that education for girls can be supported and fostered by incorporating private spaces for girls to manage

menstruation.

Menstrual hygiene management (MHM) is defined by the United Nations International Children’s Emergency Fund (UNICEF, 2014) as the “use of clean material to absorb or collect menstrual blood and this material can be changed in privacy as often as necessary for the duration of the menstrual period. MHM includes soap and water for washing the body as required, and access to facilities to dispose of used menstrual management materials” (p. 16).

In the context of this thesis, MHM is in direct reference to private spaces in which girls can wash menstrual rags and themselves. Previous studies have used small sample sizes, relying on qualitative, self-reported data to report barriers related to MHM, which include poverty, hygiene taboos, inadequate information on menstrual management, poor social support, and insufficient water, sanitation, and hygiene (WASH) facilities in schools (Marni Sommer et al., 2016; Marni Sommer, Hirsch, Nathanson, & Parker, 2015; Marni Sommer & Sahin, 2013).

The growing qualitative research suggests that adequate sanitation facilities/infrastructure in schools as it relates to MHM may improve student participation or cognitive function, particularly for girls at the onset of menstruation (Alexander et al., 2014; Long et al., 2013;

Sommer, Hirsch, Nathanson, & Parker, 2015; UNICEF, 2011); however, the quantitative literature has recently begun to explore this association. Of these studies, educational outcomes typically address absenteeism, with few studies assessing sanitation at the school level using cognitive development measured by test scores as an outcome (Grant, Lloyd, &

Mensch, 2013; Sclar et al., 2017; Shallwani, 2015; Tegegne & Sisay, 2014; Zegeye, Megabiaw, & Mulu, 2009). A lack of studies assessing sanitation, specifically the role of gendered spaces at the school level, using achievement scores as an educational outcome highlights a gap this study aims to fill.

Quantitative studies have explored the health impact of sanitation by focusing on

anthropometric outcomes or on infectious diseases (Sclar et al., 2017). However, The World Health Organization (WHO, 1978) defines health beyond disease or infirmary, as “a state of complete physical, mental, and social well-being” (p. 1). Additionally, as stated in the Ottawa

(9)

3

Charter (WHO, 1986), “Health is a positive concept emphasizing social and personal resources, as well as physical capacities… is not just the responsibility of the health sector, but goes beyond healthy life-styles to well-being (p. 1). Therefore, sanitation related to the provision of MHM spaces has the potential to address not only disease but also other aspects of well-being, such as the ability to attend school and the development of cognitive abilities, measured by academic achievement.

1.2 Study Aim

Using secondary data from The Young Lives Project, the present study aims to examine the effect of MHM spaces, on academic achievement and the extent in which environmental factors at the individual- , home-, and school/community levels can help explain those differences among Ethiopian school girls. Exploring the effect of MHM can have important policy implication and encourage national and state governments to allocate resources in an appropriate and effective manner to improve health and education, and, thus economic progress and global poverty reduction in Ethiopia.

1.3 Research Questions

The study is guided by the following research questions:

1. To what extent do adolescent girls’ math and English academic achievement differ by schools that provide MHM spaces and those that do not provide MHM spaces?

2. Is there a significant difference in academic achievement scores for individuals who attend schools with MHM spaces and those that do not provide MHM spaces while controlling for characteristics at the individual, home, and school/community?

(10)

4

2. Theoretical Framework

An ecological perspective guided the present study, which was originally proposed by Bronfenbrenner in 1977 in his human development framework (Bronfenbrenner, 1977). The ecological perspective is designed to draw attention to the dynamic interrelations among various personal and environmental factors in health (Mclaren & Hawes, 2005). In the earliest formulation of the ecological perspective, Bronfenbrenner posits that “the ecological

environment is conceived as a set of nested structures, each inside the other like a set of Russian dolls” (Bronfenbrenner, 1979, p.3). Bronfenbrenner’s metaphor suggests that an individual’s development is affected by, and affects, multiple systems/levels originally described in abstract terms of micro-, meso-, exo-, and macro- systems. These systems support and guide an individual’s development, but each system is specific to an individual’s life offering diverse possibilities of growth trajectories.

Illustrated in Figure 1, at the most intermediate level is the microsystem. Microsystems

include intrapersonal and interpersonal influences on an individual, which occur in the family, school, amongst peers, etc. This is the system in which an individual has direct contact with his or her immediate setting, and influences are strongest. The mesosystem involves the connection and interaction between two or more settings. This is the system connecting two or more environments in which an individual is a part of. It can be seen as a “system of microsystems” (Bronfrenbrenner, 1979, p. 25). For instance, the connection of time from home to school merges two settings (school and home influences) which fall in the

mesosystem. The next system is the exosystem, which an individual does not have immediate contact, such as institutional infrastructures involving politics, law, economics etc. While individuals do not have direct involvement with larger social systems, they feel the effects of what happens. For instance if a parent loses a job due to political turmoil, an individual would feel the financial repercussions in his or her life. Finally, the macrosystem refers to the

broader patterns of culture or subculture which shape ideologies and give meaning to institutions and systems, thus ultimately influences an individual’s environment

(Bronfenbrenner, 1979). Overall, Bronfenbrenner’s ecological theory of development posits that to understand an outcome, it is necessary to identify the parts related to the whole.

(11)

5

Figure 1. Bronfenbrenner’s ecological theory of development (adapted from Santrock et al, as cited in Mclaren & Hawe, 2005)

In health promotion, the ecological framework draws attention to individual and

environmental determinants of behavior at different levels of influences (Mclaren & Hawe, 2005). There are two guiding principles of the ecological perspective. The first is the

assumption of interaction and causal reciprocity among levels, underscoring the importance of interventions that are targeted and evaluated at different levels; the second being that

appropriate changes in the environments will lead to changes in the individuals (Mclaren &

Hawe, 2005). This suggests that for students to achieve academically in schools, their environment needs to be appropriate and supportive of the desired outcome. Based on this framework it is suggested that the provision of MHM spaces will result in better learning outcomes and that individual characteristics, support of family and school/community environment will play a role in academic achievement.

In applying this framework, the present study considered academic achievement as the outcome of interactions among factors at three levels— the individual-, home-, and school- level within the micro- and meso-systems. The framework is applicable to this study because it treats the interaction between factors at different environment levels with equal importance

(12)

6

and provides a comprehensive understanding of environmental factors that influence student achievement.

(13)

7

3. Literature Review

The literature used in this thesis is peer-reviewed and published in academic databases found through Oria. Searched terms include ‘water’, ‘sanitation’, ‘WASH’, ‘schools’, ‘education’,

‘academic achievement’, ‘menstruation’, ‘menstrual health management”, “MHM”,

‘Ethiopia’, and ‘health’. As MHM is an emerging topic, findings from the grey literature have also been cited.

3.1 Menstrual Hygiene Management in Schools

MHM has recently emerged as a development issue and both qualitative and quantitative studies suggest that poor MHM leads to school absenteeism, distraction, and disengagement (Hennegan & Montgomery, 2016). To begin, strong evidence from girls’ narratives regarding challenges with MHM in schools have been widely reported in the literature, particularly in low and middle income countries (LMIC) across sub-Saharan Africa, Asia, and South America (Sommer, Hirsch, et al., 2015). Feelings of fear, shame, and embarrassment in managing menstruation and a lack of guidance prior to the onset of menstruation are often reported (Mahon & Fernandes, 2010; Mason et al., 2013; McMahon et al., 2011; Sommer, 2010; Thakur et al., 2014). Studies have indicated girls lack clean, safe, and private spaces with access to water and soap to clean the body and menstrual materials (Alexander et al., 2014); adequate time to manage menstruation (Fehr, 2011); and hygienic menstrual products (Long et al., 2013; Mason et al., 2013). While there is growing qualitative evidence

suggesting the gendered impacts of inadequate WASH facilities on the participation of girls in school, quantitative studies have recently emerged studying this association.

Studies demonstrating quantitative casual associations of the challenges girls face in managing menstruation is lacking (Hennegan & Montgomery, 2016; Sclar et al., 2017; Sommer et al., 2016). Identified studies of MHM have focused on health and educational outcomes (Sumpter

& Torondel, 2013). Studies focused on health outcomes have looked at reproductive tract infection, laboratory confirmed bacterial vaginosis, or self-reported vaginal discharge using self-reported menstrual management as the exposure (Ali, Sami, & Khuwaja, 2007;

Balamurugan & Bendigeri, 2012; Younis et al., 1993). These studies show that access to improved sanitation through provision of hardware interventions, such as WASH infrastructure, can lower rate of sanitation-related illnesses, which may lead to healthier states to attend school and, therefore, better cognitive abilities. Other studies have explored the impacts of adequate

(14)

8

sanitation in schools which promotes comfortable learning environment and, therefore, a desire to attend schools. The identified studies which have assessed educational outcomes have focused on school absence, often term absenteeism, and have reported mixed results for the effect of sanitation (Dreibelbis et al., 2013; Freeman et al., 2012; Grant et al., 2013). For instance, Freeman et al. (2012) conducted a study in Kenya assessing the impact of school based WASH intervention (schools receiving hygiene promotion training and water treatment versus schools receiving hygiene promotion training, water treatment, and sanitation [provision of latrines]) on primary school children’s attendance. They found that after stratified analysis by gender, both interventions were effective in reducing absence among girls than boys compared to the control (Freeman et al., 2012). Although this study found that WASH interventions could be effective in reducing gender disparities, it does not clearly identify the mechanism by which girls benefit more, such as the role of private spaces. A study that assessed perception of private spaces with menstruation related absenteeism was conducted by Grant et al. (2013) in Malawi.

They found that perceived lack of privacy in the school toilet area was associated with 2.64 greater odds of absence during the last menstrual period (p < .05); but overall, they did not find evidence for school level variance in menstruation related absenteeism (Grant et al., 2013).

Further, a hardware intervention study exploring the effect of the menstrual cup, an improved sanitary technology, on attendance, reported a significant and negative, yet small effect that menstruation could have on attendance (Oster & Thornton, 2011). While the qualitative literature reports the many challenges that girls face in managing menstruation in school, the latter two studies (Grant et al., 2013; Oster & Thornton, 2011) cast doubt on the supposition that menstruation may be associated with absence. It is likely that absence and learning outcomes correlate, but not necessarily.

3.2 Menstrual Hygiene Management and Academic Achievement

Limited studies have explored cognitive development using measures of academic achievement as an educational outcome in assessing the influence of sanitation. As MHM relates to sanitation the following review addresses studies that address sanitation in general. Of studies assessing sanitation and cognitive development identified in a systematic review by Sclar et al. (2017), four studies investigated sanitation at the household or community level, while only one study assessed sanitation at the school level. The manner of assessment of cognitive measures varied for each study which include: child development questionnaires administered by parents, the Raven’s Progressive Matrices, Wechsler Pre-School and Primary Score Intelligence, and

(15)

9

national education reports with a majority of studies following a cross-sectional design (Cameron Manisha Olivia, Susan, 2013; Maika et al., 2013; Santos et al., 2008; Spears, 2012).

At the school level, Shallwani (2015) used Agna Khan Foundation achievement test in early primary school in Pakistan but found no association between quality of school toilets and water facilities and student achievement test scores. A lack of studies assessing sanitation, specifically the role of gendered spaces for girls, at the school level using achievement tests scores as an outcome highlights a gap in the literature this study aims to fill.

3.3 Factors Associated with Educational Outcomes

Until recently, educational outcomes have been under researched due to the absence of achievement data, but in 2000, with the introduction of the Ethiopian National Learning Assessment (ENLA), progress on understanding children’s educational outcomes in relation to wider socio-economic trends have been made (UNICEF, 2015). The following section presents influences on educational outcomes organized around how three levels of environmental influences that interact to affect children’s educational trajectories. The following section elaborates on these factors in detail and draws on the empirical literature in order to provide face validity of the variables used in the present study.

3.3.1 Individual characteristics Child health

There is extensive literature suggesting that the role of health plays a role in cognitive

development. Studies indicate that a child’s health may affect the ability to attend school, and thus, develop better cognitive abilities (Ali et al., 2007; Balamurugan & Bendigeri, 2012;

Younis et al., 1993). Other studies have reported menstrual pain as a reason for girls to be absent from school, while other studies report that the presence of disease within a school due to improper sanitation has caused parents to be reluctant to send their children to school (Colclough, Rose, & Tembon, 2000; Hennegan & Montgomery, 2016; Snilstveit et al., 2016;

Tamir, 2015). Further, stomach pains is amongst the most cited illness reported by Ethiopian students (Colclough, Rose, & Tembon, 2000). Thus, health as it relates to menstrual stomach pains, may affect attendance at school and the development of cognitive abilities and,

therefore, be related to test score results.

(16)

10 3.3.2 Family characteristics

Household wealth/parental education

Household wealth is an important factor in determining access to education as schooling incurs a range of upfront and hidden costs (Hunt, 2008). Previous studies conducted in developing countries consistently report that household wealth improves a child’s education (Admassu, 2015; Woldehanna, Ferede, Girma, Alemu, & Getachew, 2012). This relates to MHM as challenges in managing menstruation via effective menstrual management products is often tied to poverty and the ability to purchase hygienic menstrual solutions. To illustrate, due to low cost and easy availability, most menstruating girls in Ethiopia report cloth, soft paper, and rags as commonly used menstrual protection materials (Tamir, 2015). The issue around low-cost menstrual management solutions is the increased risk of menstrual leakage and stained clothing, which can affect girls’ concentration and attention in school. Of the identified studies on household associations and educational outcomes, studies have used large-scale national surveys, such as the Ethiopia Rural Household survey, and have used proxy measures of socio-economic status (Admassie & Bedi, 2003; Weir, 2011). Factors at the family level that have shown significant positive effects on educational outcomes include parental education and household wealth (Admassie & Bedi, 2003; Weir, 2011).

3.3.3 School/community characteristics:

School locality/ travel time to school

There is substantial evidence from the ENLA to support various school disadvantages effects educational outcomes. Identified school factors that have a significant effect on children’s achievement on ENLA test scores include the distance and time it takes to travel to school, and the availability and condition of school facilities (Jebena, 2013). Findings from Round 1 Young Lives school survey in Ethiopia reported that a quarter of rural schools have children that travel too far for them to stay in school for a full day, suggesting that travel time is an important cause of underachievement in certain areas (Frost & Rolleston, 2013). A study of menstrual patterns of secondary school adolescents in northwest Ethiopia concluded that better socioeconomic status for girls attending urban schools may put them at an advantage as they do not have to deal with stress that accompanies having to travel long distance to school every day, experienced by rural girls (Zegeye et al., 2009). The literature also cites that if schools are located too far parents/guardians are more reluctant to send their child to school as girls are seen as vulnerable to sexual harassment enroute (Colclough et al., 2000; Hunt, 2008).

(17)

11

With increased attention on educational attainment in relation to MHM, this study will contribute to the literature in two ways. First, with a large number of studies focusing on absenteeism, little is known about the effects of sanitation, especially MHM spaces, on cognitive development. Therefore, studying cognitive development using achievement test scores will add to the fairly, new emerging literature. Additionally, while most of the literature has focused on the home environment in relation to child development, the study aims to investigate the effect of MHM space on performance on cognitive math and English tests, and whether differences are attenuated by adjusting for individual-, family-, and school/community level characteristics.

(18)

12

4. Methods

4.1 The Young Lives Project

The current study analyzed secondary data using the Young Lives international longitudinal study of childhood poverty from the 2016-2017 Ethiopia school survey, and adopts a post- positivist paradigm in that knowledge can be produced through approximation of empirical evidence (Neuman, 2011). Unless otherwise noted, all factual information from the following section is from Young Lives (Rossiter, Azubuike, & Rolleston, 2017).

The Young Lives study is a longitudinal study of childhood poverty conducted in 4

developing countries, including Ethiopia, since 2002 and traces the lives of 12,000 children in two age groups, the ‘Younger Cohort’ born in 2001-2002 and an ‘Older Cohort’ born in 1994- 1995. Young Lives have followed these children spanning a period of 15 years, collecting data at the household and community level. In 2010 and 2012-2013, Young Lives conducted a school survey to explore a subset of the younger cohort children’s experiences of schooling and education in depth, examining issues of school quality and effectiveness at the primary level. In 2016-2017 a school survey was conducted at the upper primary level.

4.1.1 Young Lives sampling strategy

The first school survey conducted in 2010 followed the Younger Cohort children and

included 20 core sites across the regions of Addis Ababa, Amhara, Oromia, Southern Nations, Nationalities, and People's Region (SNNP), and Tigray that were purposely selected in 2001, which “ensured cultural and geographic diversity of the country, including urban-rural differences, but with a pro-poor bias and a focus on areas with food insecurity” (Outes-Leon and Sanchez, as cited in Rossiter & Azubuike, 2017, p. 6) . The second school survey in 2012-2013 included the same 20 sentinel sites but was extended to 10 additional sites in the emerging regions of Somali and Afar, following the same criteria as the original sites (see Figure 1). In 2016-2017, in collaboration with the Ethiopian Development Research Institute (EDRI), Young Lives surveyed the same 30 sites in the 2012-2013 survey, but used a census including all schools within the site’s geographic boundaries, irrespective of ownership. All Grade 7 and 8 students (Young Lives cohort children and their peers) were included in data collection. Thus, the 2016-2017 school survey sampling approach is statistically

representative of the sites sampled.

(19)

13

Figure 1. Map of 2016-2017 school survey sites (Rossiter et al., 2017)

4.1.2 Survey instruments

Survey instruments used in this study include student and school level questionnaires (includes information on math and English sections) and school facilities observations. The survey was administered in two waves, Wave 1- at the beginning and Wave 2- at the end of the school year, by trained fieldworkers from EDRI. Additionally, students’ learning levels were assessed via math and English cognitive tests at both study waves. All math and English items were pre-piloted and tested for reliability and validity by Young Lives researchers using techniques from Classical Test Theory and Item Response Theory (Azubuike, Moore, & Iyer, 2017). All students present at Wave 1 were included in the sample and followed up at Wave 2, without replacement of absentees.

4.1.3 Data quality

The data hold a high quality for several reasons. Having students re-measured on achievement at two time-points provides a comparison on how much schools contribute to a student’s progress and improvements from one testing period to the next. A second measure of children’s achievement on a subject improves the reliability and robustness measuring academic achievement as the test and re-test design provides an analytical advantage when

(20)

14

assessing learning quality, particularly in LMIC where less is known about the educational context (Iyer & Moore, 2017).

4.2 Study Sample

The sample for the following study was selected based on several criteria. As the impact of MHM is gender specific, 5958 female respondents were eligible. However, only female students enrolled in the same section/school between test waves were retained. Additionally, only female respondents considered adolescents (14 to 19 years) were kept using age as a proxy measure of menstruation, based on the literature finding that the median age of menstruation is 14.8 years (13.9-15.3) in Northwest Ethiopia and WHO’s definition of adolescents (WHO, 2001; Zegeye et al., 2009). The final sample included 3844 adolescent girls between the ages of 14 and 19 years (M = 14.86, SD = 1.02).

4.3 Study Variables 4.3.1 Dependent variables

Math and English test scores at Wave 1 and Wave 2 were the academic achievement measures used, totaling four dependent variables. Measuring English literacy and math levels is

pertinent for the study context because Ethiopia strives to be a middle-income country and improved literacy and math levels are needed in the application of science and technology, which are driving instruments of wealth (Federal Ministry of Education, 2015). Often achievement levels are used as a check against learning criteria, as seen by the Ethiopian National Learning Assessment. However, measuring how students test at one single point in time leads to substantial bias as achievement is associated with a student’s socioeconomic status and her learning to date. Additionally, with non-random sorting of students into schools and classrooms, there is risk that educational outcomes are systematically correlated with school and classroom inputs and processes. Therefore, by accounting for how much improvement a student makes from one testing period to the next, there can be less biased estimates of how much background characteristics and school factors affect a student’s learning (Azubuike et al., 2017).

Young Lives’ math tests reflect the curricular expectations for Grades 7-8 and was developed in consultation with the Ministry of Educations’ Mathematics and Science Improvement Centre. It included eight content domains: 1) basic number competency, 2) integers, rational

(21)

15

numbers, powers and bases, 3) fractions, decimals, ratios and percentages, 4) area, perimeter, volume and surface area, 5) geometry and shapes, 6) algebra, 7) measurement, charts, and graphs, 8) reasoning, problem solving, and applications in daily life.

The English language test is a reflection of transferrable skills that can be used in Ethiopia with relevance for continuing education, labor market opportunities and social mobility (Graddol as cited in Rossiter & Azubuike, 2017). English is also the language of instruction for all secondary-level subjects, therefore, assessing students’ English ability provides a good indication of future achievements. The English test covers content that upper primary grade students are expected be familiar with after learning English from Grade 1 and will need in secondary schools. The four domains of the English test include: 1) word identification, 2) word meaning and contextual vocabulary, 3) sentence construction and comprehension 4) reading/comprehension.

Both math and English tests involved 40 multiple choice items with a sub-set of items that were common in both test waves. The value labels of the each test items were 0 =

incorrect/blank and 1 = Correct. For each subject, the raw score at Wave 1 and Wave 2 was provided in the Young Lives dataset.

4.3.2 Independent variables

Independent variables were pulled from school facilities observations and student background questionnaires, which included home and family background and life experience outside of school.

4.3.2.1 Main independent variable

Menstrual hygiene management variable (MHM group): Among the various WASH measures available in the school survey, WASH measures that were gender/menstruation- specific to females were chosen. Two items fell under MHM spaces:

1) Place for girls to wash menstrual rags: “Is there a place for washing out menstrual rags (for female students close to the toilet facility)?” coded 1 = yes; 2 = no 2) Private space for girls to wash “Is there a place where female student can wash

themselves privately away from boys and male teachers?” coded 1 = yes; 2 = no

(22)

16

The two items were transformed into a categorical variables. Students were allocated to two conditions based on availability of MHM spaces at the attended school 1) None 2) MHM available (schools that provided either/or both MHM spaces); coded 0 and 1, respectively.

The survey also included an item regarding separate toilets for girls/boys: “Are there separate toilets for male and female students”, and toilet type: “what type of toilet for use by students is most common” with flush toilet, pit latrine/dry latrine, other, and no toilets as options.

However, most sampled students (96.6 percent) attend schools that provide separate toilets for girls and boys, with pit latrines/dry latrines being the prominent toilet type (90.5 percent).

Low variability of these measures makes assessing separate toilets and toilet type impractical;

therefore, these measures were not included in the study.

4.3.2.2 Individual level covariates

Age was a self-reported continuous variable captured by the question “What age are you?”

Child health was measured via several questions in the Young Lives survey. However, a proxy question of “Do you have any health problems (stomach problems) that regularly affect you in school?” was chosen as stomach problems could be interpreted to be related to menstrual pains.

This item was coded 0 = Yes, 1 = No.

4.3.2.3 Family level covariates

Household wealth was measured by questions on students’ household durable assets through the question of “Which of the following do you have at home?” with the following options:

table, chair, bed with mattress, radio, telephone, fridge, bicycle, and car or truck. Answers were coded 0= yes and 1= no. For the purpose of this study, it was decided to use ownership of a bicycle as a proxy measure of household wealth as owning a bicycle represents wealth accumulation and can serve as an income-generating asset and as a mean of transporting individuals to school (Grant et al., 2013).

Mother’s education was assessed through the question: “What is your mother’s highest level of education?” with the following categories: never been to school, Up to Grade 4, Up to Grade 8, Up to Grade 10, technical and vocational education and training (TVET) or Diploma, Up to Grade 12, University, and I don’t know. After running descriptive statistics, only a small percent

(23)

17

of the sample had mothers who had TVET or diploma (5.6 percent), attended school up to grade 12 (6.1 percent), or completed university education (4.5 percent). Thus, it was decided to collapse these levels into the one category ‘higher education or vocational training.’ Responses to ‘I don’t know’ was considered missing data. This variable was coded as follows 0 = Never been to school, 1 = basic education, 2 = general primary education, 3 =general primary secondary education, 4 = higher education or vocational training.

4.3.2.4 School level covariates

Travel time to school was a self-reported continuous variable captured by the question “How many minutes does it usually take you to get to school?”

School geographic area measured by ‘locality’, coded 0 = Rural, 1 = Urban.

4.4 Data Analysis

Statistical analyses were performed using the Statistical Package for the Social Sciences (SPSS) version 25. All analyses excluded cases pairwise with no replacement for missing data (see Section 5.1). A statistical significance level (α) of .05 was used for each statistical

analysis.

The first part of the analysis included descriptive analysis of the data and coding of variables as described above. Descriptive statistics were obtained for all variables and were presented using standard statistical parameters such as frequencies, percentages, means, and standard deviations (See Table 1 and 2 in the Results chapter). Outliers classified as ‘extreme’ by SPSS were investigated. However, outliers were retained given that they did not have a strong influence on the mean. Second, a one-way between groups analysis of variance (ANOVA) was conducted between the main independent variable (MHM group) with all dependent variables. This was followed by bivariate analyses between all dependent variables and various independent covariates at the individual-, family-, and school- level using

independent sample t-tests, Pearson Product-Moment Coefficient (r), and one-way between groups ANOVA.

(24)

18

To address the study’s first research question, a mixed between-within subject ANOVA was conducted for each outcome (math and English scores). To address the second research question, a mixed between-within analysis of covariance (ANCOVA) was conducted between the MHM variable and each subject outcome, controlling for variables (covariates) that were found to be significant in bivariate analysis.

4.5 Ethical Consideration

This thesis utilized existing data from the Young Lives school survey, so no direct contact was made with study participants. The use of secondary data saves times, money, and resources;

however, the concerns with secondary data use is the potential harm to study participants due to the varying amounts of identifying information that may be provided and the fact that the original data was not collected to address the present research questions (Tripathy, 2013). In order to use Young Lives’ secondary data, a project request was sent to UK Data Archives for non-commercial use and was approved in April 2018. The data was provided with no identifying information of participants ensuring data protection, anonymity, and confidentiality of study participants. The data was used solely for the intended purpose outlined in the project request and was under the direct supervision of researchers at the University of Bergen.

Ethical approval for the original protocol and data collection rounds of the Young Lives study was received from the London School of Hygiene and Tropical Medicine Research Ethics Committee and six other partner ethics committees including the College of Health Sciences in Ethiopia. Young Lives follows the ethics guidance of the Department of International Development, University of Oxford, Association of Social Anthropologies of the Commonwealth, and Save the Children Protection Policy (Young Lives, 2019). Additionally, Young Lives’ local in-country research and policy team of economists, educationalist, social anthropologists, developmental psychologists, epidemiologists, nutritionists, social workers, sociologists and political scientists actively discussed ethics in order to develop a shared understanding of research in the study context (Morrow, 2013). While working with local teams to address ethical challenges may reduce ‘stranger involvement’ and power dynamics between researchers and participants, it does not completely eliminate them (Morrow, 2013). It is important to recognize that social differences, including social class and social divisions of gender and age, between researchers and participants are likely to affect research participation.

(25)

19

Ethics is a fundamental part of research. In collecting data on children, Young Lives researchers obtained informed consent from parents or caregivers, and from children at each round of fieldwork (Morrow, 2013). Due to the country context, children may be taught from an early age to obey adults or authoritative figures, which makes it difficult from them to refuse to participate in a study, but Young Lives made efforts to ensure that no adverse consequences would results in non-participation. Ethical consideration was also addressed through the following provisions: 1) fieldworkers were provided with training in research ethics and provided with fieldwork manuals that contained ethics guidance 2) survey researchers and fieldworkers were required to report cases of ethical difficulties to supervisors in order resolve them (Morrow, 2013).

(26)

20

5. Results

5.1 Univariate Analysis

Descriptive analyses of all variables were conducted on the study sample to provide an overview of the data (N = 3844). Of the sample, a majority of respondents, 66.4 percent (n = 2553) attended schools that did not provide any MHM spaces, whereas 33.6 percent (n = 1291) of respondents attended schools that provided either a private space to wash menstrual rags or a private space to wash themselves.

Most of the sample attended schools in urban localities (78.5 percent), and 84.8 percent did not indicate a predisposition to stomach problems while attending school (.1 percent missing).

Less than 30 percent of respondents had mothers who completed general secondary education or obtained higher education (missing 12.8 percent). Seventy-eight point four percent of respondents did not own a bicycle, which served as an income-generating asset and as a mean of transporting individuals to school (missing .1 percent). See Table 1 for descriptive statistics of all categorical variables.

(27)

21 Table 1

Frequency of Categorical Variables, N=3844

Frequency Percent (Valid Percent)

Main independent variable MHM group

None 2553 66.4 (66.4)

MHM available 1291 33.6 (33.6)

Total 3844 100.0

Child level variables

Child health- stomach problems

Yes 579 15.1(15.1)

No 3259 84.8 (84.9)

Total 3838 (missing n=6) 99.9 (missing .1)

Family level variables

Mother's highest level of education

Never been to school 1044 27.2 (31.3)

Basic education 585 15.1 (17.4)

General primary education 753 19.6 (22.5)

General secondary education 463 12.0 (13.8)

Higher education/TVET or diploma 509 13.3 (15.2)

Total 3354 (missing n=490) 87.2 (missing 12.8)

Wealth item- Bicycle

Yes 827 21.5 (21.5)

No 3013 78.4 (78.5)

Total 3840 (missing n=4) 99.9 (missing .1)

School/community level variables Locality

Rural 827 21.5 (21.5)

Urban 3017 78.5 (78.5)

Total 3844 100.0

In the study sample, Wave 1 math and English scores ranged from 2 to 40, with a mean of 16.48 (SD = 6.17) and 19.44 (SD = 6.73), respectively. Wave 2 math scores ranged from 2 to 40 (M = 19.08, SD = 7.03) while English scores ranged from 5 to 40 (M = 20.42, SD = 6.85).

Time to travel to school ranged from one minute to 180 minutes. The mean travel time was 24.94 minutes (SD = 20.41). See Table 2 for descriptive statistics of continuous variables.

(28)

22 Table 2

Descriptive Statistics of Continuous Variables

Skewness Kurtosis

N Missing Min Max Mean

Std.

Deviat ion

Statis tic

Std.

Error Stati

stic

Std.

Error Math- Wave 1 3844 0 2 40 16.48 6.17 0.62 0.04 0.27 0.08 Math- Wave 2 3222 622 2 40 19.08 7.03 0.41 0.04 -0.44 0.08 English- Wave 1 3726 118 2 40 19.44 6.73 0.42 0.04 -0.32 0.08 English- Wave 2 3346 498 5 40 20.42 6.85 0.28 0.04 -0.48 0.08

Age 3844 0 14 19 14.86 1.02 1.25 0.04 1.27 0.08

Minutes it takes

to get to school 3812 32 1 180 24.94 20.41 2.08 0.04 7.18 0.08

When reviewing the distribution of the data in order to meet the assumption of normality, it was found that the skewness and kurtosis values for achievement scores deviated from the normal value of 0 (see Table 2). However, an inspection of the shape of the distribution of scores for each achievement test showed scores to be reasonably distributed (see Figures 2-5 in appendix).

5.2 One-Way Between Groups ANOVA: Effect of MHM Group on Achievement Scores

One-way between groups ANOVA was conducted to explore the impact of the MHM group on academic achievement, as measured by Wave 1 and Wave 2 math and English test scores.

Participants were divided into 2 groups according to the availability of MHM spaces at their schools (Group 1: None, Group 2: MHM available). See Table 3 for the means and standard deviations for each of the academic achievement variables by MHM group.

(29)

23 Table 3

Descriptive Statistics of One-Way ANOVA for the Effect of MHM Group on Achievement Scores

95% CI

N M SD SE LL UL Min Max

Math- Wave 1 None 2553 16.34 6.23 0.12 16.10 16.58 2 40 MHM available 1291 16.74 6.06 0.17 16.41 17.07 3 37 Total 3844 16.48 6.17 0.10 16.28 16.67 2 40 Math-Wave 2 None 2192 18.88 6.94 0.15 18.59 19.17 2 40 MHM available 1030 19.51 7.20 0.22 19.07 19.95 5 40 Total 3222 19.08 7.03 0.12 18.84 19.33 2 40 English- Wave 1 None 2475 18.94 6.75 0.14 18.68 19.21 2 40 MHM available 1251 20.41 6.59 0.19 20.05 20.78 5 38 Total 3726 19.44 6.73 0.11 19.22 19.65 2 40 English- Wave 2 None 2232 20.13 6.86 0.15 19.85 20.42 5 40 MHM available 1114 21.00 6.79 0.20 20.60 21.40 5 38

Total 3346 20.42 6.85 0.12 20.19 20.65 5 40

Note. CI = confidence interval; LL = lower limit; UL = upper limit

Welch’s F test was used as Levene’s test indicated a violation of assumption of homogeneity of variance for Wave 2 Math scores, (p < .05) (See Table 4). The one-way ANOVA of students’ average score on Wave 2 math test revealed a statistically significant main effect, indicating there is a difference between the MHM groups on average mean scores, Welch’s F(1, 1950) = 5.54, p = .02.

(30)

24 Table 4

Levene’s Test of Homogeneity of Variances

Levene

Statistic df1 df2 Sig.

Math- Wave 1 Based on Mean 0.00 1 3842 .947

Based on Median 0.00 1 3842 .988

Based on Median and with adjusted df 0.00 1 3831 .988

Based on trimmed mean 0.02 1 3842 .887

Math- Wave 2 Based on Mean 5.94 1 3220 .015

English- Wave 1 Based on Mean 0.01 1 3724 .905

English- Wave 2 Based on Mean 0.02 1 3344 .882

One-way ANOVA results for the remaining test scores indicated that student’s average score on each measure of academic achievement revealed a statistically significant main effect, indicating there is a significant difference between MHM groups on average mean scores, however the calculated effect sizes using eta squared were very small (see Table 5).

Table 5

One-Way Between Groups ANOVA Summary Table for the Effect of MHM Group on Achievement Scores

SS df MS F p η²

Math- Wave 1 Between Groups 137.26 1 137.26 3.60 .054 .00 Within Groups 146341.73 3842 38.09

Total 146478.99 3843

English- Wave 1 Between Groups 1801.14 1 1801.14 40.19 <.001 .01 Within Groups 166881.54 3724 44.81

Total 168682.68 3725

English- Wave 2 Between Groups 551.59 1 551.59 11.792 <.001 .00 Within Groups 156416.39 3344 46.78

Total 156967.99 3345

(31)

25 5.3 Bivariate Analysis of Covariates

5.3.1 Independent-samples t-tests

Independent samples t-tests were used to compare all waves of mathematic and English academic achievement test scores with the following dichotomous variables: child health (stomach problems), wealth item (ownership of bicycle), and school locality.

Starting with child health, there was a significant difference in Wave 1 English scores for individuals who reported stomach issues (M = 18.72, SD = 6.51) from those who did not (M

= 19.58, SD = 6.76), t (3718) = -2.79, p < .01, two-tailed. A significant difference is also seen in Wave 2 English scores for those who reported stomach problems (M = 19.58, SD = 6.70) from those who did not report stomach problems (M = 20.57, SD = 6.86), t (3339) = -3.02, p <

.001 (two-tailed). While the difference in means were significant (wave 1 mean difference = -0.86, 95% Cl: -1.46 to -0.26; wave 2 mean difference = -.99, 95% Cl: -1.64 to -.35, two- tailed) both findings represented a small effect (d = 0.13 and 0.15). There was no statistically significant difference in scores for individuals who reported stomach problems and those who did not have stomach problem for math achievement scores. See Table 6 for more details.

Table 6

T-Tests: Achievement Score Differences Between Groups That Reported or Did Not Report Stomach Problems

Yes- stomach problems

No- stomach

problems 95 % CI

M SD M SD df t p LL UL

Cohen's d Math - Wave 1 16.67 6.07 16.45 6.19 3836 0.79 .430 -.033 0.77 0.04 Math - Wave 2 18.86 6.81 19.13 7.07 3214 -0.76 .450 -0.94 0.42 0.04 English- Wave 1 18.72 6.51 19.58 6.76 3718 -2.79 .010 -1.46 -0.26 0.13 English- Wave 2 19.58 6.70 20.57 6.86 3339 -3.02 <.001 -1.63 -0.35 0.15

On average, individuals who owned a bicycle (M = 20.44, SD = 6.71) scored significantly higher on English achievement tests at Wave 1 than those who did not own a bicycle (M

=19.17, SD = 6.94; t(3720) = 4.75, p < .001). The mean difference between groups was 1.27 scores higher (95% Cl: .75 to 1.79) with a small effect size (d = 0.19). The relationship was also significant for Wave 2 English achievement tests. Those who owned a bicycle at Wave 2 had a mean of 21.67 (SD = 6.96) while individuals who did not own a bicycle had a mean of

(32)

26

20.08 (SD = 6.78; t(3342) = 4.75, p < .001). The effect size was large (d = 0.23) with a mean difference of 1.59 (95% Cl: 1.02 to 2.15, two-tailed). There was no statistical significance of for Wave 1 and Wave 2 math test score means. See Table 7 for more details.

Table 7

T-Tests: Achievement Score Differences Between Groups That Reported Ownership of Bicycle or Non-Ownership of Bicycle

Owns a bicycle

Does not own a

bicycle 95 % CI

M SD M SD df t p LL UP

Cohen's d Math - Wave 1 16.47 6.28 16.48 6.15 3838 -0.08 .940 -0.49 0.46 0.00 Math - Wave 2 19.27 7.35 19.03 6.94 3217 0.80 .420 -0.35 0.84 0.03 English- Wave 1 20.44 6.71 19.17 6.94 3720 4.75 <.001 0.75 1.79 0.19 English- Wave 2 21.67 6.96 20.08 6.78 3342 5.50 <.001 1.02 2.15 0.23

Differences in math and English academic achievement test scores were significant across all test waves for school locality. The magnitude of the differences was moderate for mathematic scores at Wave 1 (mean difference = .22; 95% CI: -3,60 to -2,79; d = 0.56) and Wave 2 (mean difference= -. 26; 95% CI: -4.40 to -.034; d = 0.59), but was large for English scores at Wave 1 (mean difference = -.86; 95% CI: -6.45 to -5.65; d = 1.06) and Wave 2 (mean

difference = -.99; 95% CI: -6.56 to -5.68; d = 1.26). See Table 8 Table 8

T-Tests: Achievement Score Differences Between Groups That Attend Schools Located in Rural Versus Urban Locality

Rural Urban 95 % CI

M SD M SD df t p LL UL

Cohen's d Math - Wave 1 13.97 5.00 17.16 5.00 1614 -15.35 <.001 -3.60 -2.79 0.56 Math - Wave 2 16.05 6.07 19.92 7.05 1251 -14.34 <.001 -4.40 -0.34 0.59 English- Wave 1 14.46 4.59 20.74 6.64 1828 -29.77 <.001 -6.45 -5.65 1.06 English- Wave 2 15.65 4.91 21.77 6.72 1590 -27.36 <.001 -6.56 -5.68 1.26

(33)

27

5.3.2 Pearson’s product-moment correlation coefficients

The relationship between math and English achievement scores with continuous variables (age and time traveled to school) were investigated using Pearson’s product-moment r. There was a weak, negative correlation between age and Wave 2 mathematic achievement scores (r

= -.05, n = 3222, p < .01) with higher scores associated with lower ages. Time to travel to school was a significant correlate of academic achievement across both subjects and all test waves, with higher scores associated with shorter travel time to school. The relationship between age and Wave 1 math scores and age with Wave 1 and Wave 2 English test scores were not significant correlates. See Table 9 for details.

Table 9

Pearson Product-Moment Correlations Between Continuous Variables and Academic Achievement Measures

Math English

Variable Wave 1 Wave 2 Wave 1 Wave 2

Age -0.03 -0.05* -0.02 -0.03

Travel time to school -0.07* -0.08* -0.10* -0.13*

*. Correlation is significant at the 0.01 level (2-tailed)

5.3.3 One way-between groups ANOVA

A one-way- between groups ANOVA was used to explore differences in means of academic achievement for each dependent variable between groups based on mother’s education. The independent variable represented the mother’s education level: 1) never been to school 2) basic education 3) general primary education 4) general secondary education 5) and higher education/TVET or diploma. Notably, average means for both subject and test wave increased as the level of mother’s education increased. See Table 10 for the means and standard

deviations for each of the 5 groups.

Exploring the Effect of Menstrual Hygiene Management Spaces on Academic Achievement: Evidence From Young Lives' School Survey in Ethiopia

Acknowledgements

Table of Contents

Abstract

Acronyms and Abbreviations

1. Introduction

2. Theoretical Framework

3. Literature Review

4. Methods

5. Results